The bahnaric-phoneme project is driven by the objective to empower Bahnaric language speakers, fostering communication within their community and with other ethnic groups. Our overall goal is to contribute to the empowerment and connectivity of Bahnaric ethnic communities through targeted advancements in speech processing.
The data utilized in this project originates from a broader research initiative spearheaded by Prof. Quan Thanh Tho. Please ensure to obtain his consent prior to using the data for any other purposes.
- This project leverages widely-used acoustic features, including MFCC, F0, and energy.
- In order to address the challenge of varying speech signal lengths, we employ a sliding window approach to divide the speech signals into frames. Subsequently, we compute the average of each feature within each frame.
- Install the required packages by running
pip3 install -r requirements.txt. - Generate the features by running
python3 src/dataset.py. - Train the model by running
python3 src/train.py. - Evaluate the model by running
python3 src/evaluate.py.
- Install docker desktop for windows. guide
- Choose
Rebuild and Reopen in Containerin the popup window.